4 Cautions That Truly Matter When Building A Chatbot

★ 3 min read ★

 

 

Chatbots are changing the way people interact on the Internet. According to a survey by Oracle that analyzed 800 decision makers, including chief marketing officers, chief strategy officers, senior marketers, and senior sales executives, 80% of the respondents said they already implemented chatbots in their business or planned to use chatbots by 2020. Therefore, people have the decision of whether to enter and win the race to chatbot adoption or get left behind in the new era. However, before you jump into building a chatbot, there are certain things to keep in mind that actually make a difference.

1. Define clearly the goal of your chatbot

This seems to be such an obvious thing, yet so many people fail to do this properly. Building a chatbot doesn’t start with the first line of code, it starts when you define what you want the chatbot to do. Chatbots have their own strengths and weaknesses. They beat humans when it comes to math abilities or data retrieval skills, but they do poorly with context-rich questions or descriptive qualitative answers from users. More often than not, people would think “I want a bot that can do this”, but it would turn out a different bot, or even a different solution, would solve the root cause of the problem better.

What are the problems your business wants to solve? What does your business need? What function will chatbot help to perform better or even replace it? Make sure you have the right answers to these questions before moving on to the next steps.

2. Beware of the invisible UI

First impressions count.

A good UI is the best chance a chatbot has to convince users to start chatting with it. However, different from websites or applications that have blocks of text content, images and buttons to invite users to interact, chatbot’s main interface is a text-based conversation. Thus, it’s important to offer cards, buttons, and other graphic elements for interactivity and ease of use. You should never let users second-guess what the chatbot can and cannot do.

Also, give your chatbot a name and a face. Humans by nature look for humans in other things, that’s why we keep seeing human faces in random objects. Chatbots are no exception. How users perceive your chatbot greatly impacts how they engage and interact with it. Don’t let your chatbot pretend to be a human though. Set user expectations right by ensuring that people know when they’re chatting with a chatbot.

3. Consider time available for users

When users chat with your chatbot, it shouldn’t take them more time to complete their intentions than when they navigate themselves. No one would like to be asked 10 different questions before getting an answer to their inquiry. Long blocks of texts are also a big no-no, not only they make the conversation less natural and bulky, but users are less likely to read them as well, considering human attention span nowadays is approximately 8 seconds, which is 1 second shorter than a goldfish’s.

That’s why it’s important to take user behavior into consideration when mapping the chatbot’s workflow. This is where it gets tricky, but also where value is added by the chatbots, since reducing time not only create a better experience for users but also help to cut down costs.

4. Train with care

Chatbots are only as good as the training they are given. Training chatbots involves setting up intents and loading them with expressions. You should stay away from creating too specific intents and try to use entities to interpret the topic. In most cases, it makes more sense to set up a global intent and use entities to define users’ requests, rather than using very specific intents that can overlap and cause confusion.

Another thing is that the point of entities is to extract key information to turn into coding language. Therefore, avoid tagging every single word in a sentence as an entity. Besides, also take precautions to add only necessary expressions and not to overflow intents with new user expressions and mess up your existing training.

In a nutshell, with these 4 points in mind, you’re on track to build a chatbot that can attract, deliver and evolve over time.

Leave a Reply

Your email address will not be published. Required fields are marked *

one × 5 =